api_health
ChatGPTCheck whether the configured PHOTO iD API and its database are healthy.
assign_smart_tag_task
ChatGPTSet who a Smart Tag task is assigned to. Replaces the current assignees. Pass an empty list to unassign everyone. Use profile ids from list_potential_assignees. Choose people using their availability, presence, and relationship (contact, subordinate, company). Confirm with the user before changing assignees. Requires mcp:write.
create_assignment
ChatGPTCreate a PHOTO iD assignment owned by the authenticated user. Use when the user wants a new claim/assignment from text fields (name, address, claim numbers, damage details). Extract values from the prompt or from documents the client has already read — this tool does not upload files. Required: claim_insured_name (insured / contact display name). The assignment is always created for the OAuth user; you cannot create for another profile. Requires the mcp:write scope. Leave external_id unset unless the user explicitly asks to set the PHOTO iD integration / external ID. Claim, policy, carrier, vendor, and order numbers from FNOL or chat go in claim_number or claim_policy_number — never in external_id. If the user names a category/workflow template (Settings > Templates), pass template_name. Match is case-insensitive among templates the user can access, including PHOTO iD defaults. Omit it to use the default folder tree. To schedule a visit for the authenticated user (same as the web app self-appointment on create), pass appointment_at. Omit it to create the job unscheduled. appointment_at must be in the future. Prefer ISO 8601 with an offset (e.g. 2026-09-12T14:00:00-05:00). Naive local times (2026-09-12T14:00:00) need appointment_timezone (IANA, e.g. America/Chicago). If the offset is present, timezone is optional. appointment_at in the result is Unix epoch milliseconds as a string. After a successful create, include web_url in the reply as a markdown link (label it as the assignment in the PHOTO iD web app) so the user can open the job. Do not invent a URL if web_url is missing.
create_personal_quick_captions
ChatGPTCreate reusable quick captions for the authenticated user only. Call only when the user directly asks to create personal quick captions or explicitly confirms a prior recommendation. Before recommending new ones, call list_quick_captions and exclude captions already saved (personal or company). Never create company quick captions or imply that this tool can do so. Requires mcp:write. Each caption must be concise, reusable across assignments, and no longer than 250 characters.
create_smart_tag_task
ChatGPTCreate a Smart Tag task on an assignment, optionally assigned to contacts. Requires mcp:write. Call list_potential_assignees first when choosing people. Prefer confirming the assignee and due date with the user. Types: repair, inspection, assessment, documentation. Priorities: low, medium, high, urgent. due_date is YYYY-MM-DD. Newly assigned people are notified and granted viewer access if they are not already on the assignment.
find_assignments
ChatGPTFind PHOTO iD assignments the authenticated user can access. Use this to resolve an assignment_id before image/label tools when the user refers to an insured name, claim number, or external ID instead of a raw id. Prefer claim_number or external_id when available (more unique). If multiple matches are returned, ask the user which assignment to use. At least one of q, claim_insured_name, claim_number, or external_id is required. Requires the mcp:read scope.
flag_assignment_photo_issues
ChatGPTSave QA findings on an assignment (replaces the previous review). Use after get_assignment_photo_review when you have spotted mistakes the inspector made. Kinds: misplaced_image, incorrect_caption, missing_caption, incorrect_smart_tags, missing_smart_tags, other. Use other only for another human documentation mistake (for example a photo of the wrong property, or a near-duplicate shot filed twice) — never for software or data problems such as duplicate ids, missing urls, or sync errors. Severities: low, medium, high. For misplaced_image set image_id, current_leaf_path, suggested_leaf_path. For caption issues set suggested_caption as a concise caption (one short sentence or noun phrase; longer only if the user asked for more detail). An empty findings list clears the saved review. This records flags for the user; it does not move photos or change captions. To move photos, use organize_assignment_images after the user confirms. Requires mcp:write. Confirm with the user before overwriting a prior review.
get_assignment_images
ChatGPTLoad assignment photos as MCP image blocks for visual inspection. Pass a short list of image_ids (from list_assignment_images or get_assignment_photo_review). Returns, in order, a text line per photo (image_id and filename) followed by that photo's image block. Failed photos become a text line for that id; the rest still return. Metadata-only tools do not include pixels — use this tool when you need to see photos.
get_assignment_photo_review
ChatGPTLoad an assignment so you can QA the inspector's photo work. Use when the user asks to review, audit, or flag problems on a job. Review what a person got wrong while documenting the inspection: photos filed under the wrong category (leaf_path), captions that do not match the photo, missing captions, and missing or wrong Smart Tags. This is not a technical audit. Do not look for or report software or data problems (duplicate or malformed ids, missing urls, sync or upload errors, odd timestamps, schema oddities). If you notice something like that, tell the user in chat; never save it as a finding. Returns assignment context (damage type / description), the category tree, each photo's metadata (public url, caption, leaf_path, Smart Tags), and any findings already saved. This tool does not include image pixels. Call get_assignment_images with the image_ids you need to see; do not open the public urls yourself. Then call flag_assignment_photo_issues with what you found. If the user asks you to put photos in the right folders, use organize_assignment_images after confirming. Do not rewrite captions or tags unless the user asked to fix them — flag first. If you do write or suggest a caption, keep it concise by default. When the user wants organize, caption, and/or Smart Tag work, plan from this response (and get_assignment_images as needed) before any mcp:write calls. If they did not say whether to overwrite existing captions or Smart Tags, ask whether to fill only missing/empty values, fix incorrect ones, or overwrite broadly. Before applying writes, confirm with brief counts in one short summary, for example: "2 uncaptioned images to caption. 5 smart tags to apply across 3 images. 8 images to reorganize." Wait for approval unless they already agreed to that exact plan. Requires mcp:read. Resolve assignment_id with find_assignments if needed.
get_image_labels
ChatGPTGet caption and Smart Tags for one assignment photo, with the image attached. Captions are short report text under the photo. Smart Tags are structured pins (type, severity, label, notes, x/y). Use this to review or prepare corrections. When writing a new caption, keep it concise by default. The tool result includes an MCP image block for this photo when it can be loaded from PHOTO iD storage.
list_assignment_images
ChatGPTList photos on a PHOTO iD assignment (metadata only). Returns each image's public url, caption (text under the photo), category leaf_path, and smart_tag_count. This tool does not include image pixels. Use get_assignment_photo_review when the user wants a QA pass (wrong category, bad captions), or this list when you need ids and metadata. Call get_assignment_images with the image_ids you need to see; do not open the public url yourself.
list_potential_assignees
ChatGPTList people you can assign to Smart Tag tasks. Includes contact info and availability. Returns approved contacts, subordinates, teammates from companies you own or admin, and yourself. Each person includes name, email, relationship groups, live presence, usual working days/hours, and calendar exceptions in the coming days. Use this before assigning a task. Pick someone whose schedule and role fit the due date and work; do not invent profile ids.
list_quick_captions
ChatGPTList reusable quick captions the user can apply to photos. Returns the authenticated user's personal captions and any company quick captions from companies they belong to. Use this optionally while writing image captions when saved wording would help. Prefer an exact saved caption when it accurately describes the image. Do not force a saved caption onto an image it does not fit. If you identify useful reusable captions that are not already in this list, recommend creating them as personal captions by saying: "I recommend creating the following reusable quick captions: [...]. Would you like me to create them for you?" Do not create them until the user confirms. Company captions are read-only here.
list_smart_tag_tasks
ChatGPTList Smart Tag tasks on an assignment. Tasks are follow-up work linked to one or more Smart Tags. Use this to see existing titles, due dates, and current assignees before creating or reassigning work.
organize_assignment_images
ChatGPTMove photos into the correct category on a PHOTO iD assignment. Use when the user asks to organize, file, or put misplaced photos in the right folder. Prefer categories from get_assignment_photo_review (full leaf_path like "Dwelling > Interior"). Do not call until the user approved the planned moves. When asking for confirmation, include a brief count in the same summary as any caption or Smart Tag writes on that job (for example: "8 images to reorganize."). This does not change captions or Smart Tags. Requires mcp:write. At most 25 photos per call.
ping
ChatGPTHealth check that the PHOTO iD MCP server is running.
summarize_assignments
ChatGPTSummarize the authenticated user's PHOTO iD assignments. Use this for a conversational version of the Assignments dashboard: status counts (in progress, inspection complete, report completed), next scheduled visits, overdue appointments, unscheduled work, and recently created assignments. Prefer this over find_assignments when the user asks what is on their plate, what is coming up, or for an overview — not when looking up one claim by name or number. Counts match the web app header: in_progress is status 040, inspection complete is 050, report completed is 060. Upcoming is soonest future appointment; unscheduled has no active appointment. appointment_at and creation_date are Unix epoch milliseconds as strings — convert them to local dates when talking to the user. Requires the mcp:read scope.
update_image_caption
ChatGPTUpdate the caption (text under a photo) on an assignment image. Use for photo descriptions / report text — not for Smart Tags. Default to a concise caption: one short sentence or noun phrase naming what is shown (place plus visible subject or damage). Skip filler, hedging, and restating the category or Smart Tags. Write a longer caption only when the user asks for more detail. When the user asks to use their saved/custom quick captions, call list_quick_captions first and use an exact saved caption (personal or company) when it fits the image. Do not create quick captions as a side effect of updating an image. After captioning, if new wording would be useful across future assignments, compare it with the saved list and say: "I recommend creating the following reusable quick captions: [...]. Would you like me to create them for you?" Create them only after the user confirms, and only as personal captions. Requires mcp:write. Do not call until the user approved the planned writes. If they did not say whether to overwrite existing captions, ask whether to caption only empty/missing photos, fix incorrect captions, or rewrite all. Treat a non-empty caption as existing. When asking for confirmation, give brief counts in one summary with any organize or Smart Tag work on the same job (for example: "2 uncaptioned images to caption. 5 smart tags to apply across 3 images. 8 images to reorganize."). Each call replaces the full caption on one image.
upsert_smart_tags
ChatGPTCreate or update Smart Tags (structured pins) on an assignment photo. Merges into existing tags by id; omit id to create a new tag. Required per tag: type, severity, label. Types: damage, humidity, structural, electrical, plumbing, hvac, roofing, flooring. Severities: low, medium, high, critical. Optional: description, x/y (0-100, default 50), scale (0.5-2). Requires mcp:write. Do not call until the user approved the planned writes. If they did not say whether to change existing Smart Tags, ask whether to add only missing tags, update incorrect tags (send each tag's id), or re-tag photos from scratch (new tags without id; existing pins stay unless you update them by id). When asking for confirmation, give brief counts in one summary with any organize or caption work on the same job (for example: "2 uncaptioned images to caption. 5 smart tags to apply across 3 images. 8 images to reorganize.").
whoami
ChatGPTReturn the PHOTO iD identity and scopes authorized for this MCP session.